II - NORMAL CIRCUMSTANCES

For dams as for other industries, under normal circumstances safety is essentially based on human behaviour, and risk assessment should pay a great deal of attention to this point.

IIa - First-filling failures

First-filling failures most often occur during the first flood season after completion, but some have happened during construction or a few years afterwards. For ungated dams, the highest ever reservoir level may be reached-without crest overtopping-during a flood after 50 or 100 years of incident-free operation, but this can have a substantial impact on gravity-dam stability or the risk of piping in the upper part of some fill dams.

First-filling failures (including piping failures during the first 2 years of operation) are as follows (outside China):

Fill dams caused over 50% of failures but less than 10% of victims, probably due to the time time taken for full breaching to occur.

Statistically. the dams most at risk have been old masonry dams and the more recent buttress and multiple-arch dams. The safest have been concrete gravity dams.

Since 1950, 0.15% of dams (outside China) have failed at first filling. Height does not seem to improve safety: of 2,000 dams higher than 60 m, 3 have failed (Malpasset in 1959. Teton in1975, and Guhuo in China in 1993), and 2 reservoirs caused accidents in 1961. When there was a landslide into Vaiont dam reservoir, the overtopped 200 m high dam did not fail but the wave drowned over 2,000 people weIl downstream (night-time failure). An induced earthquake in Koyna reservoir did not endanger the 100 m high dam but caused 100 victims close to the reservoir.

Of 20 failures of concrete or masonry dams, 15 were caused by the foundation (usually sliding or piping) and 5 by masonry weakness (gravity or buttress dams). None was caused by poor concrete quality.

Few fill dams built after 1965 have failed at first filling and their failures seem to be due to foundations or embedded pipes (progress in filters and drains seems to have been effective). Some gate accidents have been dangerous.

Due to general technical progress and monitoring, the future rate of first-filling failures may be in the range of 1‰ but reducing it further may be more difficult: foundation conditions will never be perfectly known; design errors (including in key details) may occur; it is not impossible for a full lift to be of poor quality in homogeneous fill dams or RCC dams; landslides or induced earthquakes are not easily assessed. Moreover, better monitoring could not have predicted all past failures, many of which (St Francis, Vaiont, Malpasset, Teton) were unexpected.

Of the 10 or 20,000 dams to be built in populated areas within the next 50 years, some are bound to fail. Traditional risk assessment is generally based on physical or statistical data: it could be more effective to study the human causes: foundation failures are due not to geological conditions but to lack of knowledge about them, or to ill-adapted design, and most past failures in masonry dams were due not to masonry but to the work of masons.

Most industries take great pains over the study of human causes of accidents. For dams, these human causes may be lack of experience of designers or contractors (but also overconfidence of competent ones), lack of communication, ill-defined responsibilities, control and specifications looking closely at details and overlooking key points, lack of operator training, ill-adapted emergency planning, etc. The probability of accidents is more intimately linked with these human problems than with the physical data normally taken into account in the design.

IIb - Ageing (including piping)

Thirty-five failures due to ageing have been reported (25 involving dams built before 1950). Yearly rates of failure were about 2 x 10^-4 before 1950 and have been 2 x 10^-5 since.

Piping in the foundation has caused a number of fill dams and some masonry or concrete dams to fail. Failures due to piping in the body of fill dams or to embedded pipes are not uncommon. Few masonry or thin concrete structures and no thick concrete structures have failed.

The progress of safety has been due to progress in design and construction but possibly even more to maintenance and monitoring, and particularly to proper visual inspection and careful follow-up of increases in leakage. Monitoring has prevented many failures and reduced the consequences of others: the average number of victims to ageing failure is 5 times less than for other causes.

The probability of failure may be more closely linked to the quality of monitoring than to physical data. Long fill dams need properly adapted solutions (half of the fill dams that have failed in the last forty years were more than 1,000 m long).

IIc- Construction accidents

The main human risk is often overlooked, and occurs in normal circumstances during construction : it is work accidents during construction. According to ICOLD Bulletin 80, "Dam construction sites: accident prevention", this human risk is greater than the total risk from dam failure and the corresponding cost has been in the range of 3% of direct construction costs, i.e. more than the cost of all failures combined.

Appropriate care is the duty of contractors, and especially of site management, but also of owners and consultants (contractual safety clauses) and supervisory entities. In industrial countries the rate of accidents has been halved in 15 years (1970-1985).

Risk assessment and practical methods for improvement are presented in ICOLD Bulletin 80 which is available in ICOLD Office.

Spending less than 0.5% of the total construction cost for safety measures not only avoids loss of lives and permanent disabilities but may also save between 1 and 2% of construction costs.

Previous < 1 2 3 4 > Next